When Manus first dropped, it had the field to itself. Consumer agents didn’t exist as a category yet. Cursor was an editor. ChatGPT couldn’t run a browser. Manus showed up with a working autonomous loop and a clean handoff between tools, and for a few weeks the demos were the only honest thing the agent space had going for it.
That was the high point.
The story since has been choreography. Founded in China by a small team, relocated to Singapore as the regulatory weather shifted, then acquired by Meta. You can already feel the new owner in the product. UI choices stopped being weird and started being predictable. The grays got a touch bluer. The edges got rounder. Personality drains out the middle the way it always does.
And now, per the latest reporting, the acquisition is barred from closing on the China side.
What I keep coming back to is the integration. Manus shipped one of the few agent products where the seams between planning, browsing, executing, and reviewing don’t make me want to do the work myself. That is a harder problem than most teams will admit. The shelf is full of “agent” products that are really demos in a trench coat. Manus had a system.
Whether the system survives a half-finished acquisition is the actual question. I’m rooting for it. Not because I have any love for the way Meta tends to absorb good ideas, but because the alternative is everyone in this space flattening into the same wrapper around the same three labs.
Use it while it’s still recognizable.